20674
Assessing the Use of Eye-Blinking As a Measure of an Individual's Engagement with Ongoing Visual Content

Thursday, May 14, 2015: 11:30 AM-1:30 PM
Imperial Ballroom (Grand America Hotel)
C. Ranti1, W. Jones1, G. J. Ramsay1, A. Klin1 and S. Shultz2, (1)Marcus Autism Center, Children's Healthcare of Atlanta and Emory University School of Medicine, Atlanta, GA, (2)Department of Pediatrics, Marcus Autism Center, Children's Healthcare of Atlanta, Emory University, Atlanta, GA
Background: Reduced social engagement is a pervasive, early-emerging feature of Autism Spectrum Disorder (ASD). Yet research in this area has been neglected, largely due to the absence of measures capturing an individual’s subjective perception of stimulus salience. This study provides a methodological solution, building upon an approach that quantifies engagement by analyzing patterns of eye-blinking among groups of viewers. Given that blinking results in a loss of visual information, viewers unconsciously modulate blink-timing: they are least likely to blink during moments perceived as most salient. While this provides insight into what is engaging to a group, quantifying what an individualperceives as engaging presents challenges. For instance, individuals blink relatively infrequently – thus, moments of engagement (indexed by statistically significant blink inhibition) are more readily quantified within a group. Overcoming this challenge is critical: individual metrics of engagement can provide much-needed insight into the subjective experience of a child with ASD.

Objectives: Determine whether meaningful information can be recovered from an individual’s eye-blinking, by examining: 1) the conditions under which eye-blinking can be used to classify individuals by experimental group during a highly-controlled pilot task; 2) whether patterns of eye-blinking can be used to classify individuals by diagnostic group (TD or ASD) during a free-viewing task.

Methods: Pilot: Eye-tracking data were collected while 21 typical adults watched videos alternating between scenes of water animals and scenes of land animals. The categories of content were differentially engaging to viewers: half of the participants were instructed to count the number of water animals, and the other half, land animals. Natural Viewing:Eye-tracking data were collected while school-age TD children (n=48) and children with ASD (n=48) watched scenes of social interactions.

Results: Pilot: An SVM classifier correctly labeled 95% of participants by experimental group, using eye-blink timing (Figure 1). Classification was robust over varying scene durations and over viewers with very low or high blink rates. Natural Viewing: To determine when TD and ASD viewers were most engaged, we identified moments of blink inhibition for a subset of each diagnostic group (“training groups”), using permutation testing. We then calculated blink rates of the remaining TD and ASD viewers (“test groups”) during moments perceived as highly engaging by the training groups. As predicted, TD test viewers blinked less during moments perceived as engaging by the TD training group than during moments that ASD viewers perceived as engaging. ASD test viewers showed the opposite pattern (Figure 2). However, individualswithin the test groups did not show this pattern.

Conclusions: Pilot task analyses demonstrate that individual patterns of eye-blinking can be used to classify participants according to the timing of their engagement. In the naturalistic paradigm, different moments were identified as engaging to the TD and ASD training groups. Critically, two independent samples of TD viewers (TD train and TD test) showed similar patterns of eye-blinking, as did the independent samples of ASD viewers. Future analyses will explore the conditions under which individual patterns of eye-blinking during naturalistic viewing can be used as a diagnostic classifier.